1 is 
ral 
8) 
the 
  
ate 
nt, 
1C€ 
tial 
tial 
Yueqin Zhou 
  
  
  
  
  
  
Figure 5. Subsidence map from levelling Figure 6. Geocoded D-interferogram 
measurements (96.01-97.10) 
Even though the quality of the differential interferogram is degraded by noise due to low coherence because the time 
difference between the two passes is too big (21 months), we can still see clear fringes, especially in hard surface areas. 
Each fringe represents 2.8 cm of slant range change along the line of sight. A similar pattern can be seen by comparison of 
Figure 5 and Figure 6. 
4.4 Error modelling 
In order to estimate the errors between the height change measured 
by D-InSAR and by levelling, the levelling points must be projected 
into the D-InSAR height change model (Figure 7). This was done 
using the following method. Firstly, through correlating the geocoded 
SAR intensity image and the georeferenced orthoimage of the same 
area, some feature points are identified. Then the coordinate 
transformation can be done using the coordinates of corresponding 
points. Finally, the coordinates of the levelling points at the height 
change model can be calculated through coordinate transformation. 
The cross marks shown in Figure 7 indicate the levelling points used 
as the control points for error modelling (total 124 points). Another 
226 levelling points are also projected into the height change model, 
and are used to evaluate the error modelling. 
  
   
    
Fi . Geocoded height ch 1 
Based on the assumption of (6), the height change differences igne zi Geocodediheishtchange mode 
between the D-InSAR measurements and the levelling measurements contain two parts, i.e. systematic errors modelled by 
(7), and random errors. With the height change differences at the control poin ts, the coefficients of (7) are estimated by least 
square method. The coefficients of the polynomial are selected through significance test. Figure 8 gives the histogram of the 
residuals at the control points. The residuals at the 226 checkpoints are also calculated, leading to the histogram shown in 
Figure 9. 
  
  
30 60 
  
  
504 
204 
40- 
30- 
104 204 
    
  
  
  
  
  
  
  
  
  
  
Std. Dev= 1.46 404 Std. Dev= 1.21 
Mean = -.18 Mean = -.03 
0 ke N = 124.00 0 : _|N = 226.00 
Es252:2:55838282323 9$ $99 999594299259 
Sm STR ST ONE NG EN 
Unit:cm Unit: cm 
Figure 8. Histogram of residuals at control points Figure 9. Histogram of residuals at checkpoints 
  
International Archives of Photogrammetry and Remote Sensing. Vol. XXXIII, Part Bl. Amsterdam 2000. 357